Before offering generative AI to the public, complete applicable security assessments and algorithm filings, and label generated content
A public generative-AI service capable of shaping opinion or mobilizing people must undergo the required security assessment and algorithm filing, including updates or cancellation when it changes or closes. Generated images and videos need labels. Private tools not offered publicly fall outside these requirements. Penalties come through laws such as the Cybersecurity Law.
What it takes
No filing fee, but China's cyberspace-authority process takes time. A personal tool not offered to the public does not trigger this framework.
What you may gain
Providers of generative-AI services with public-opinion attributes or social-mobilization capability must conduct security assessments under national rules and complete algorithm filing, changes and cancellation under algorithm-recommendation regulations. Those terms mean the ability to influence public views or organize people. Providers must label generated images, videos and other relevant content. The Chinese interim measures took effect August 15, 2023 and apply nationwide.
Context & considerations
The measures do not specify fine amounts themselves; penalties depend on the Cybersecurity Law, Data Security Law and Personal Information Protection Law, hence grade B. Original Supreme Court website texts for civil cases involving AI-generated copyright or portrait-right infringement were not verified for this chapter and are not cited.
Research & references
Cyberspace Administration of China and six other departments (2023). Interim Measures for the Management of Generative Artificial Intelligence Services, Articles 12 and 17. https://www.gov.cn/zhengce/zhengceku/202307/content_6891752.htm